{"slug": "ryan-s-reads-ai-sycophancy", "title": "Ryan's Reads: AI Sycophancy", "summary": "Two new studies reveal that sycophantic AI responses make users less likely to apologize or change behavior, and over time reduce satisfaction in human relationships. Researchers found AI is 50% more likely to agree with users than humans, leading to increased dependence and decreased prosocial intentions.", "body_md": "# Ryan's Reads: AI Sycophancy\n\nBoy... I knew AI being sycophantic was a problem, but it hadn't really occurred to me how much of a problem it could be until I read a couple papers I saw recommended on LinkedIn today.\n\nThis all feeds into what will be discovered only in the long term to be harmful (at the individual level) effects of AI use. I'm [concerned about it with education](https://www.schuetzler.net/blog/what-im-reading-ais-effects-on-education/), and now I'm also more concerned about it in interpersonal relationships.\n\n## Decreases prosocial intentions and promotes dependence\n\nCheng, M., Lee, C., Khadpe, P., Yu, S., Han, D., & Jurafsky, D. (2026). Sycophantic AI decreases prosocial intentions and promotes dependence. *Science*, *391*(6792), eaec8352. [https://doi.org/10.1126/science.aec8352](https://doi.org/10.1126/science.aec8352?ref=schuetzler.net)\n\n- Uses a series of very interesting tests, including YTA posts from reddit.com/r/AmITheAsshole to test just how sycophantic AI really is - turns out it's about 50% more likely to tell you you're right than a human. These are cases where it's an overwhelming majority against the storyteller, and GenAI is all in on telling them they're right anyway\n- Once told they're right, people are less likely to apologize or take steps to mend a relationship, or to change their behavior.\n- And it turns out people like being told that they're right, and are more likely to trust the sycophantic AI, and return to use it again.\n\n## Makes human interaction feel less satisfying\n\nIbrahim, L., Hafner, F. S., Cheng, M., Lee, C., Anselmetti, R., Willer, R., Rocher, L., & Yang, D. (2026). *Sycophantic AI makes human interaction feel more effortful and less satisfying over time* (arXiv:2605.07912). arXiv. [https://doi.org/10.48550/arXiv.2605.07912](https://doi.org/10.48550/arXiv.2605.07912?ref=schuetzler.net)\n\n- Longitudinal study - 3 weeks of interactions rather than the single interaction from the Cheng paper above\n- People want AI that \"gets\" them - really, what they want is an AI that centers them and doesn't challenge them to think about others perspectives. The answers people like from AI are more sycophantic and self-centered, supporting the users own self-image.\n- Over time, this erodes satisfaction in human relationships - they're too hard, and less likely to satisfy your selfish desires.", "url": "https://wpnews.pro/news/ryan-s-reads-ai-sycophancy", "canonical_source": "https://www.schuetzler.net/blog/ryans-readai-sycophancy/", "published_at": "2026-06-09 15:41:08+00:00", "updated_at": "2026-06-19 17:13:55.556136+00:00", "lang": "en", "topics": ["ai-safety", "ai-ethics", "large-language-models", "generative-ai", "natural-language-processing"], "entities": ["Cheng", "Lee", "Khadpe", "Yu", "Han", "Jurafsky", "Ibrahim", "Yang"], "alternates": {"html": "https://wpnews.pro/news/ryan-s-reads-ai-sycophancy", "markdown": "https://wpnews.pro/news/ryan-s-reads-ai-sycophancy.md", "text": "https://wpnews.pro/news/ryan-s-reads-ai-sycophancy.txt", "jsonld": "https://wpnews.pro/news/ryan-s-reads-ai-sycophancy.jsonld"}}